A likelihood‐based sensitivity analysis for publication bias on the summary receiver operating characteristic in meta‐analysis of diagnostic test accuracy

Abstract

In meta‐analysis of diagnostic test accuracy, the summary receiver operating characteristic (SROC) curve is a recommended method to summarize the diagnostic capacity of a medical test in the presence of study‐specific cutoff values. The SROC curve can be estimated by bivariate modeling of pairs of sensitivity and specificity across multiple diagnostic studies, and the area under the SROC curve (SAUC) gives the aggregate estimate of diagnostic test accuracy. However, publication bias is a major threat to the validity of the estimates. To make inference of the impact of publication bias on the SROC curve or the SAUC, we propose a sensitivity analysis method by extending the likelihood‐based sensitivity analysis of Copas. In the proposed method, the SROC curve or the SAUC are estimated by maximizing the likelihood constrained by different values of the marginal probability of selective publication under different mechanisms of selective publication. A cutoff‐dependent selection function is developed to model the selective publication mechanism via the t$$ t $$‐type statistics or P$$ P $$‐value of the linear combination of the logit‐transformed sensitivity and specificity from the published studies. It allows us to model selective publication suggested by the funnel plots of sensitivity, specificity, or diagnostic odds ratio, which are often observed in practice. A real meta‐analysis of diagnostic test accuracy is re‐analyzed to illustrate the proposed method, and simulation studies are conducted to evaluate its performance.

DTAmetasa: an R shiny application for meta-analysis of diagnostic test accuracy and sensitivity analysis of publication bias

“…16,17 In many published metaanalyses of DTA, the presence of PB was tested by the asymmetry of the funnel plot on the log-transformed diagnostic odds ratio (lnDOR), which suggested that the -statistic of the lnDOR might be responsible for the underlying selective publication. Zhou et al 10 proposed the cutoff-dependent selection function to model different selective publication mechanisms and extended the likelihood-based sensitivity analysis for PB in meta-analysis of intervention studies 17 into meta-analysis of DTA. The method of Zhou et al adopted a general class of selection functions on the linear combination of logit-transformed sensitivity and specificity, which included the lnDOR as the special case.…”

“…The method of Zhou et al introduced the marginal selection probability, = (publish), as the sensitivity parameter, which implies the expected proportion of the published studies from the population of all studies. Given a range of values of , the SROC curves and the corresponding SAUC accounting for PB could be estimated from the conditional log-likelihood of the observed data (equation 13 in Zhou et al 10 ). Under the flexible selective publication mechanisms driven by the contrast ( 1 , 2 ) and a range of 's, the impact of PB could be assessed by the changes of the estimated SROC curves and SAUC.…”

“…Some sophisticated statistical methods have been proposed recently. 7,8,9,10 These methods adopted selection models to characterize the underlying selective publication process, and they reported good performance in adjusting PB under certain publication mechanisms. However, implementing these methods requires users to have programming skills and some specific statistical backgrounds.…”

“…DTAmetasa supports the meta-analysis of DTA as well as the sensitivity analysis of PB. Recently, Zhou et al 10 proposed the sensitivity analysis method to evaluate the impact of PB on the SROC curve or the SAUC givenbui a range of marginal selection probabilities and different selective publication mechanisms which are induced by the -type statistics of the combination of sensitivity and specificity. DTAmetasa simplifies the process of analyses and presents the results of sensitivity analysis of PB via the interactive plots.…”

“…In Section 2, we briefly describe the sensitivity analysis method of Zhou et al 10 and present the main functionalities of DTAmetasa. In Section 3, we use the real-world meta-analysis of DTA to illustrate how to implement DTAmetasa and interpret the results from sensitivity analysis.…”